Dynamic Contrast-Enhanced Breast MRI using Flexible Radial Undersampling with Compressed Sensing Reconstruction

Introduction: Flexible radial imaging allows multiple image sets, each having a different spatiotemporal balance, to be retrospectively reconstructed from the same dataset [1,2]. One of the applications that may benefit from this flexibility is dynamic contrast-enhanced breast imaging, in which the optimal spatiotemporal balance for image diagnosis is unknown. Images from radial undersampling have good overall image contrast at high temporal resolutions, but they suffer from undersampling streak artifacts that degrade image quality. Compressed sensing (CS) L1-penalized reconstruction has been shown to reduce such streak artifacts [3-5] by enforcing sparsity in the image. Here, we implement flexible golden-angle radial sampling [1,2] on a healthy human volunteer, and reconstruct images with CS over a range of spatiotemporal resolutions. We compare the golden-angle method with other sampling schemes (bit-reversed and random sampling), and assess the ability of CS reconstruction to reduce streak artifacts in each sampling method.